EP0685086B1 - Device for automatically generating a knowledgebase for an expert diagnostic system - Google Patents

Abstract

Powerful expert systems for breakdown diagnosis often decisively increase the availability of technical systems by making a diagnostic expert's knowledge available to a layman. The process meets the following requirements: for every variant in a technical system composed of modules, a knowledgebase can efficiently be drawn up. An expert system it to make it possible rapidly to diagnose every variant. According to the process, on obtaining knowledge, knowledge modules are produced i.e. one knowledge module for each type of actual module which occurs in at least one variant of the technical system. The knowledge module contains all diagnosis-relevant knowledge on the actual type of model in such a way that it applies to every example of the type. The device of the invention automatically generates a variant-specific knowledgebase in that it copies, using the current configuration of the variants, each knowledge module as much as there are examples of the actual type of module concerned and the copies are then suitably combined. It inserts the copies of the sub-modules into the module copy for a composite module.

Description

The present invention relates to a device for creating the knowledge base of an expert system for technical systems, which can occur in different variants.

State of the art

Whenever a technical system does not perform its intended function or does so insufficiently, a technical diagnosis is required on this system. The technical diagnosis consists in determining the cause of the observed misconduct and then taking a remedy that eliminates the cause. For complex technical systems, diagnosis requires the knowledge of an experienced human expert. But these are expensive and often not available. Therefore, a way must be found to make their knowledge available without using it themselves.

A modern solution to this requirement is to provide an expert system for technical diagnosis. High-performance expert systems are able to significantly increase the availability of technical systems and reduce operating costs.

An expert system today is typically constructed as a knowledge-based system, i.e. H. as a software system in which the knowledge of the human expert is stored in a structured manner in a knowledge base. The remaining components of the expert system are logically separated from the knowledge base and therefore reusable, they are collectively called the expert system shell. The expert system shell includes the problem-solving component, which carries out a diagnosis and evaluates the knowledge base, the explanatory component, which explains the actions of the expert system to the user, and the user interface. A knowledge-based system is created by selecting an expert system shell and filling the knowledge base with the structured knowledge of the human expert.

• o Die Wissensbasis eines symptom-basierten Expertensystems enthält das Erfahrungs-Wissen des menschlichen Experten über die Störungen am technischen System und über die Zusammenhängen zwischen diesen Störungen und den während einer Diagnose ermittelbaren Symptomen.
• o Die Wissensbasis eines modell-basierten Expertensystems enthält einerseits ein Modell, das die Struktur und das funktionale Zusammenwirken der Module des zu diagnostizierenden technischen Systems beschreibt, andererseits eine Beschreibung des Soll-Verhaltens und des Verhaltens im Störungsfalle aller Module.

Diagnostic expert systems are classified into symptom-based and model-based systems:

• o The knowledge base of a symptom-based expert system contains the experience-based knowledge of the human expert about the faults in the technical system and the relationships between these faults and the symptoms that can be determined during a diagnosis.
• o The knowledge base of a model-based expert system contains on the one hand a model that describes the structure and the functional interaction of the modules of the technical system to be diagnosed, on the other hand a description of the target behavior and the behavior in the event of a fault in all modules.

A model-based expert system can only be created if you can create a sufficiently precise model for the structure and the functional interaction of the modules of the technical system to be diagnosed. In many applications, this is too complex or completely impossible, because the interior of the technical system cannot be described precisely enough. In this case, only the symptom-based approach remains.

The classic and often used approach to formalize the experience of an expert is to structure the knowledge in rules of the form "FALLS premise THEN conclusion". The premise of each of these rules is a single symptom or a logical combination of symptoms that can be observed in the technical system; the conclusion is a statement about which disorders can or should be present when the premise is fulfilled and which can be excluded. This structuring z. B. "EX-SYS" and "NexpertOBJECT", two tools for building expert systems.

Another often used approach is the troubleshooting tree. The knowledge base contains a fault tree, that is a directed graph, the nodes of which stand for the possible faults in the technical system and the edges for the causal relationships below them. The leaves represent the possible causes. The knowledge base also contains connections between disorders and symptoms. The expert system "navigates" in this troubleshooting tree with the aim of reaching a page. To do this, it evaluates each symptom obtained in order to move from one node to another. The "testbench" expert system shell uses this procedure, for example.

Knowledge acquisition has proven to be a bottleneck in the creation of symptom-based expert systems; this is the process by which the experience knowledge of a human diagnostic expert is queried or otherwise captured and brought into the structure in which the expert system bases the knowledge can evaluate. Knowledge acquisition is often expensive, time-consuming and prone to errors.

The bottleneck and cost factor of acquiring knowledge is particularly important if you want to provide expert systems for all variants of a class of technical systems and only need one expert system that diagnoses a frequently used technical system. A class of technical systems often has a wide variety of variants: In order to adapt a technical system to the wishes of the customer, it is typically composed of modules. Alternative or optional modules are selected so that their interaction provides the desired functionality. Even with a relatively small number of alternatives or options, you have a huge number of possible variants: With only 10 modules with 3 alternatives each, there are already 3 ^ 10 = 59,049 possible combinations.

So far, no method was known of how to efficiently create knowledge bases in the case of a large number of variants. It is very inefficient to create a separate knowledge base for each variant "by hand" or to adapt a knowledge base "by hand" to a variant. In addition, there is often not enough experience knowledge for a particular variant.

If you want to create a single knowledge base that is valid for each variant and can therefore be used universally, you have to make explicit reference to the different variants in structured knowledge. The behavior and thus the knowledge about a module of the technical system often depends on the version of another module. When acquiring knowledge, all possible versions must be taken into account - their behavior can differ considerably, especially in the event of a fault. Case distinctions for the different variants of the technical system are required, and due to the huge number of Combination options you have to make many case distinctions. The structured knowledge becomes extremely extensive. It can hardly be overlooked, is difficult to maintain and validate and requires a lot of storage space.

A remedy is proposed in EP-A 0 352 514: a knowledge base is set up which contains the knowledge which is common to all technical systems to be diagnosed (standard rulebase). In addition, a knowledge base is developed for each variant, which contains the special features of the variant (calibration rulebase). This procedure also does not solve the problem of the high level of knowledge acquisition in most cases: For many more complex technical systems, the universal knowledge base (standard rulebase) is extensive and difficult to maintain, since it describes the entire technical system and all of its modules. If the variants of the technical system differ greatly, the knowledge base for the special features of a variant (calibration rulebase) is inevitably extensive.

• o Die Nachteile einer universell einsetzbaren Wissensbasis wurden bereits genannt
• o Ist ein technisches System zu diagnostizieren, so wird vor der eigentlichen Diagnose die Wissensbasis "von Hand" aus Bestandteilen zusammengesetzt und damit auf die jeweilige Variante zugeschnitten. Der Mensch, der diese Tätigkeit vornimmt, muß selber dafür Sorge tragen, daß sich die Bestandteile zusammenpassen. Diese Tätigkeit ist zeitraubend, kostenträchtig und nur von Spezialisten zu erledigen, da sie gute Kenntnisse der Struktur der Wissensbasis erfordert.
• o Dem Expertensystem wird für jede Variante eine eigene Wissensbasis vorab zur Verfügung gestellt. Ist ein technisches System zu diagnostizieren, so wählt das Expertensystem die passende Wissensbasis aus und verwendet sie. Das Expertensystem kann nur die Varianten diagnostizieren, für die es eine Wissensbasis besitzt; man muß also vorhersehen, welche Varianten möglicherweise einer Diagnose bedürfen werden. Andere Varianten kann das Expertensystem überhaupt nicht diagnostizieren. Bei jeder neu zum Einsatz gelangenden Variante muß eine neue Wissensbasis ausgeliefert werden. Die Menge der Wissensbasen erfordert großen Speicherplatz.

An expert system often has to be able to diagnose several or even all variants of the technical system - for example when the expert system is used in a vehicle workshop and many different vehicle variants have to be repaired. The previously known procedures have considerable disadvantages:

• o The disadvantages of a universally applicable knowledge base have already been mentioned
• o If a technical system is to be diagnosed, the knowledge base is assembled "by hand" from components before the actual diagnosis and thus tailored to the respective variant. The person who carries out this activity must ensure that the components fit together. This activity is time-consuming, costly and can only be done by specialists because it requires a good knowledge of the structure of the knowledge base.
• o The expert system is provided with a separate knowledge base for each variant in advance. If a technical system is to be diagnosed, the expert system selects the appropriate knowledge base and uses it. The expert system can only diagnose the variants for which it has a knowledge base; one must therefore foresee which variants may need to be diagnosed. The expert system cannot diagnose other variants at all. A new knowledge base must be delivered for each new variant that is used. The amount of knowledge bases requires large storage space.

A technical system often changes in the course of its use: one module is replaced by another that differs from the first one in terms of diagnosis. So far, no method was known of how to efficiently update the knowledge base of an expert system for this technical system. The previously known methods require that the entire knowledge base is searched for mentions of the “old” module and that the knowledge dependent on the “old” module is changed “by hand”.

• o die auch dann einen zeit- und kostensparenden sowie wenig fehleranfälligen Wissenserwerb für eine Klasse modularer technischer Systeme gestattet, wenn die Klasse eine hohe Variantenvielfalt besitzt,
• o und mit der man bei diagnoserelevanten Änderungen an einer Variante die Wissensbasis schnell und wenig fehleranfällig an die Änderung anpassen kann.

The object of the invention was therefore to create a device

• o which also enables time and cost-saving and less error-prone acquisition of knowledge for a class of modular technical systems if the class has a large variety of variants,
• o and with which one can adapt the knowledge base quickly and less prone to errors to the change in the case of diagnosis-relevant changes to a variant.

• o wie ein Expertensystem jede mögliche Variante des technischen Systems schnell diagnostizieren kann Sicherheitsanalysen und die FMEA ("failure mode and effect analysis") sind zwei Beispiele für Aufgabenstellungen, in denen ein Fehlermodell ("Fehlerbaum") eines technischen Systems TS benötigt wird. Im Falle hoher Variantenvieifalt treten die gerade genannten Probleme, eine Wissensbasis zu erstellen, auch in dieser Aufgabe auf. Eine zusätzliche Aufgabe der Erfindung war es daher, ein Verfahren und eine Einrichtung zu schaffen, das auch für eine Klasse modularer technischer Systeme hoher Variantenvielfalt ein vollständiges Fehlermodell zu erzeugen gestattet.

Building on this facility, a method had to be developed

• o how an expert system can quickly diagnose every possible variant of the technical system Safety analyzes and FMEA ("failure mode and effect analysis") are two examples of tasks in which a fault model ("fault tree") of a technical system TS is required. In the case of a high variety of variants, the problems just mentioned to create a knowledge base also occur in this task. An additional object of the invention was therefore to create a method and a device which also allows a complete error model to be generated for a class of modular technical systems with a large number of variants.

Solution idea

The idea of the device according to the invention is as defined in claim 1: the knowledge of the expert required for a diagnosis is divided up during the formalized storage, in a manner analogous to the modularization of the technical system. The knowledge of a module type M-type, which occurs in a variant of the technical system at least once, is structured in such a way that it applies to all copies of the type M-type. All knowledge about M-type required for diagnosis is summarized and saved in a logical module after acquiring knowledge. This logical module is referred to below as the "knowledge module". The knowledge base for a variant is created automatically by configuring knowledge modules based on the current configuration of the variant. Figure 1 illustrates this idea.

Brief description of the figures

1 illustrates the idea of automatically configuring a knowledge base from knowledge modules

FIG. 2 shows the relationships under the terms “units”, “systems”, “components”, “subsystems” and “technical system”.

3 illustrates the terms and relationships of the semantic network that structures a knowledge base that is automatically configured according to the inventive method.

Fig. 4 illustrates the three possible types of a relation between a unit type and a disturbance

Figure 5 illustrates the two possible types of relation between a role in a system type and a fault

FIG. 6 shows how a computer Wb-Creator automatically creates a complete fault map for a technical system. The current configuration of the technical system is stored in the memory Konf, WM contains all the necessary knowledge modules.

Fig. 7 shows how, during the generation of the complete fault map, the fault graph module for one unit is inserted into the SYS fault graph module at the Ro location.

8 illustrates the insertion shown in FIG. 7.

9 shows how a computer Diag diagnoses a technical system by processing a knowledge base generated according to the invention

10 shows how a result Resu is evaluated during a diagnosis

11 shows how the explanations for resu are evaluated during the evaluation of a result resu

FIG. 12 demonstrates how the relationships between results and causes, which are required for a diagnosis according to the procedure shown in FIG. 6, are obtained.

Figure 13 illustrates how knowledge is taxonomically organized.

FIG. 14 shows how the information about the current configuration of a technical system is obtained automatically from information that is assigned to unit types.

description

• o Der Speicher Konf enthält die aktuelle Konfiguration der zu diagnostizierenden Variante, d. h. die Information, aus welchen Modulen die gerade betrachtete Variante besteht und von welchem Modul-Typ jedes Modul ist.
• o Der Speicher WM enthält alle benötigten Wissens-Module. Für jeden Modul-Typ, das in mindestens einer Variante des technischen Systems auftritt, ist das zugehörige Wissens-Modul im Speicher WM enthalten. Außerdem ist in WM ein Wissens-Modul für das technische System als ganzes enthalten, das beschreibt, wie die anderen Wissens-Module zusammenzufügen sind.

The essential information for the device according to the invention is stored in two memories:

• o The memory Konf contains the current configuration of the variant to be diagnosed, ie the information of which modules the variant under consideration consists of and of which module type each module is.
• o The WM memory contains all the necessary knowledge modules. For each module type that occurs in at least one variant of the technical system, the associated knowledge module is contained in the WM memory. WM also contains a knowledge module for the technical system as a whole, which describes how the other knowledge modules are to be put together.

The knowledge modules in WM are created during the course of knowledge acquisition. A knowledge module in WM is always used for knowledge base generation if the variant for which a knowledge base is to be created has the real module type occurring at least once. The information in Konf is specific to the variant to be diagnosed, while the knowledge modules can be reused for any other variant of the technical system.

The knowledge base for a variant is created by configuring knowledge modules: A computer Wb creator to create a knowledge base determines the current configuration of the variant by reading Konf and then creates a copy of the knowledge module for each real module in the variant the type to which the real module belongs. If a real module occurs n times in the variant, n copies of the knowledge modules are created. Wb creators automatically assemble all copies of the knowledge modules generated in this way and store them in a memory Wb.

• o Das technische System besteht aus Modulen, Module haben Teilmodule usw. Die für die jeweilige Diagnoseaufgabe kleinsten austauschbaren Einheiten werden als elementare Module aufgefaßt und im folgenden "Komponenten" genannt. Für alle zusammengesetzten Module wird der Oberbegriff "Teilsysteme" benutzt. Für die Teilsysteme und das technische System als ganzes wird der Oberbegriff "Systeme" verwendet. Für Systeme und Komponenten wird der Oberbegriff "Einheiten" benutzt. Fig. 3 zeigt den Zusammenhang unter diesen Begriffen.
• o An Einheiten treten Störungen auf. Der Begriff "Störung" bezeichnet jede Art von fehlerhaftem Verhalten oder fehlerverursachendem Zustand der Einheit.
• o Störungen stehen in kausalen Zusammenhängen untereinander. Ein kausaler Zusammenhang ist von der Form "Störung A ist Folge von Störung B". Die kausalen Beziehungen unter Störungen bilden den Störungsgraph für das technische System.
• o Eine Störung, die nicht Folge einer anderen Störung ist und daher ein Blatt des Störungsgraphen ist, wird als Einfach-Ursache bezeichnet. Ziel einer Diagnose ist, die aktuell am technischen System vorliegende Einfach-Ursache zu identifizieren - oder die aktuelle Mehrfach-Ursache, die aus mehreren gleichzeitig aufgetretenen Einfach-Ursachen besteht.
• o Jeder Einfach-Ursache ist eine Abhilfe zugeordnet, das ist eine Maßnahme oder eine Folge von Maßnahmen, die die Einfach-Ursache beseitigen.
• o Um im Verlaufe einer Diagnose herauszufinden, welche Ursache am technischen System vorliegt, werden Untersuchungen duchgeführt, und zwar entweder automatisch vom Expertensystem oder von einem Menschen.
• o Die möglichen Ergebnisse einer Untersuchung heißen Resultate. Das Soll-Resultat einer Untersuchung ist das Resultat, das die Untersuchung beim störungsfreien technischen System hat.
• o Um ein während einer Diagnose gewonnenes Resultat auswerten zu können, ist es logisch mit Störungen verbunden. Zwei Arten von Relationen zwischen einer Störung und einem Resultat werden unterschieden:
  • Eine Störung "erklärt" ein Resultat, wenn die Störung für das Resultat verantwortlich gemacht werden kann.
  • Eine Störung "ist unvereinbar mit" einem Resultat.

Before it is described how the knowledge modules are structured and how they are automatically put together, the result of the joining is described, that is, what structure has an automatically generated knowledge base. The information in the knowledge base is organized as a semantic network. The nodes of this network stand for terms that are important for the diagnosis of the technical system, the edges for connections between these terms. These terms and relationships are explained below.

• o The technical system consists of modules, modules have sub-modules, etc. The smallest interchangeable units for the respective diagnostic task are understood as elementary modules and are called "components" in the following. The generic term "subsystems" is used for all assembled modules. The generic term "systems" is used for the subsystems and the technical system as a whole. The generic term "units" is used for systems and components. Fig. 3 shows the relationship under these terms.
• o Malfunctions occur on units. The term "malfunction" means any type of faulty behavior or condition causing the unit.
• o Disruptions are causally related to one another. A causal relationship is of the form "disorder A is the result of disorder B". The causal relationships among faults form the fault graph for the technical system.
• o A fault that is not the result of another fault and is therefore a sheet of the fault graph is called the simple cause. The aim of a diagnosis is to identify the single cause currently present in the technical system - or the current multiple cause, which consists of several simple causes that occurred simultaneously.
• o A remedy is assigned to each simple cause, that is a measure or a sequence of measures that eliminate the simple cause.
• o In order to find out in the course of a diagnosis which cause is in the technical system, examinations are carried out, either automatically by the expert system or by a human.
• o The possible results of an investigation are called results. The target result of an examination is the result that the examination has with the fault-free technical system.
• o In order to be able to evaluate a result obtained during a diagnosis, it is logically linked to faults. There are two types of relations between a disturbance and a result:
  • A fault "explains" a result if the fault can be held responsible for the result.
  • A disturbance "is incompatible with" a result.

These two relations are mutually exclusive. Often there is no relation between a disturbance and a result.

FIG. 2 illustrates the terms and relationships of the semantic network that structures a knowledge base that is automatically configured according to the inventive method.

• o Die Informationen über einen Einheits-Typ müssen so im Wissens-Modul strukturiert sein, daß sie nicht davon abhängen, wie eine Einheit dieses Typs verwendet wird.
• o Die Kopien der Wissens-Module, die der Rechner Wb-Ersteller erzeugt, müssen automatisch aufgrund der Informationen in den Speichern WM und Konf zusammengefügt werden.
• o Für das technische System als ganzes muß es ein Wissens-Modul geben, das festlegt, wie die Wissens-Module für die Teilsysteme zusammenzufügen sind. Zweckmäßigerweise hat dieses Wissens-Modul die gleiche Struktur wie eines für ein Teilsystem-Typ.
• o Das Wissens-Modul für einen System-Typ darf nicht davon abhängen, von welchen Typen die Untereinheiten eines Exemplars des Systems-Typs sind, da sonst das Problem der Variantenvielfalt die Erstellung der Wissens-Module ineffizient macht. Das Wissens-Modul muß aber die Folgen beschreiben, die eine Störung an einer Untereinheit hat.
• o Das Wissen über ein Teilsystem muß sich genauso aus Wissens-Modulen konfigurieren lassen wie die Wissensbasis für das technische System selbst.

In order for knowledge to be acquired efficiently and for a knowledge base to be configured automatically from knowledge modules, the organization of the information contained in the knowledge modules must meet the following requirements:

• o The information about a unit type must be structured in the knowledge module so that it does not depend on how a unit of this type is used.
• o The copies of the knowledge modules that the computer Wb-Creator creates must be put together automatically based on the information in the WM and Konf memories.
• o For the technical system as a whole there must be a knowledge module that defines how the knowledge modules for the subsystems are to be put together. This knowledge module expediently has the same structure as one for a subsystem type.
• o The knowledge module for a system type must not depend on which types the subunits of an instance of the system type are, otherwise the problem of the diversity of variants makes the creation of the knowledge modules inefficient. However, the knowledge module must describe the consequences that a fault has on a subunit.
• o The knowledge of a subsystem must be configurable from knowledge modules as well as the knowledge base for the technical system itself.

In order to meet these requirements, there is a knowledge module for the technical system as a whole and a knowledge module for each subsystem and component type that occurs at least once in the technical system. To put it briefly: there is one knowledge module per unit type. Each knowledge module is stored in a data structure "knowledge module" in the memory WM.

• o ein Störungsgraph-Modul für E-Typ, der aus allen Störungen, die an E-Typauftreten können, und den kausalen Zusammenhängen unter ihnen besteht,
• o alle Abhilfen an E-Typ, die Störungen an E-Typ beseitigen,
• o alle Untersuchungen, die E-Typ gelten,
• o für jede Untersuchung deren Soll-Resultat und deren Nicht-Soll-Resultate.

When acquiring knowledge, all knowledge modules are created and saved. The knowledge module for a unit type E-type summarizes all diagnosis-relevant knowledge about E-type. Are saved

• o a fault graph module for E type, which consists of all faults that can occur on E type and the causal relationships among them,
• o all remedies on E-type that eliminate faults on E-type,
• o all examinations that apply to the E type,
• o For each examination, their target result and their non-target results.

"A fault is saved" means that a unique identifier for the fault is entered in the memory.

• o St wirkt unmittelbar von Ein nach außen (und ist damit logische Außen-Schnittstelle)
• o St wird unmittelbar von außen an Ein bewirkt (und ist damit logische Außen-Schnittstelle)
• o St beeinträchtigt Ein intern (und ist damit keine logische Außen-Schnittstelle)

Fig. 4 illustriert diese drei Relationsarten.The knowledge module for a unit type contains a fault graph module for the unit type. The overall fault graph for the technical system is generated by automatically merging copies of these fault graph modules. To make this possible, some faults are identified as logical external interfaces in the fault graph module for the unit type. The faults that represent logical external interfaces are differentiated according to their direction of action for the unit type. There are therefore three types of relations between a disturbance St and a unit type On:

• o St acts directly from on to the outside (and is therefore a logical external interface)
• o St is immediately switched on from the outside (and is therefore a logical external interface)
• o St impaired an internal (and is therefore not a logical external interface)

Figure 4 illustrates these three types of relations.

The fault graph module for a component type contains only faults on the component type itself, not on other unit types. The fault graph module for a system type shows which faults result in a fault in a subunit of the system type. But it must not explicitly name a version of the subunit. This is the only way to meet the requirements just outlined.

This is made possible by introducing the term "role in the system type". Each sub-unit of the system type fulfills a specific role for the functioning of the system. Just as actors occupy the roles in a play, it is said that the subunits of a system occupy the roles in the system.

• o St wirkt unmittelbar von Ro nach außen (und ist damit logische Außen-Schnittstelle)
• o St wird unmittelbar von außen an Ro bewirkt (und ist damit logische Außen-Schnittstelle)

Fig. 5 illustriert diese Zusammenhänge.In the knowledge module, the description of how a subunit can be disturbed is attached to the role that the subunit occupies. Depending on its direction of action, a fault St can be connected to a role Ro in a system type S type in two ways:

• o St acts directly from Ro to the outside (and is therefore a logical external interface)
• o St is effected directly from the outside on Ro (and is therefore a logical external interface)

5 illustrates these relationships.

The connection between disturbances at different roles in S-type is described in the knowledge module for S-type, that between different disturbances at roles in S-type, however, not in the knowledge module for S-type, but in the knowledge module for the cast of Ro in S type.

The concept of "roles in a system type" is a key to handling the variety of variants of a modular system: Typically, many variants of the system type differ only in different occupations, but not in the interaction of the roles. Since the fault graph module only refers to roles, but not to subunits, it applies to every version of the system type that has this role.

• o Folge einer Störung, die unmittelbar von außen an S-Typ bewirkt wird, ist entweder eine Störung, die S-Typ intern beeinträchtigt, oder eine, die unmittelbar von außen an einer Rolle in S-Typ bewirkt wird, oder eine, die unmittelbar von S-Typ nach außen wirkt.
• o Folge einer Störung, die S-Typ intern beeinträchtigt, ist entweder eine andere Störung, die S-Typ intern beeinträchtigt, oder eine, die unmittelbar von außen an einer Rolle in S-Typ bewirkt wird, oder eine, die unmittelbar von S-Typ nach außen wirkt.
• o Welche Folge eine Störung hat, die unmittelbar von S-Typ nach außen wirkt, ist nicht im Störungsgraph-Modul von S-Typ gespeichert, sondem in einem Störungsgraph-Modul eines System-Typs, in dem ein System vom Typ S-Typ eine Rolle besetzt.
• o Folge einer Störung, die unmittelbar von einer Rolle in S-Typ nach außen wirkt, ist entweder eine Störung, die S-Typ intern beeinträchtigt, oder eine, die unmittelbar von außen an einer anderen Rolle in S-Typ bewirkt wird, oder eine, die unmittelbar von S-Typ nach außen wirkt.
• o Welche Folge eine Störung hat, die unmittelbar von außen an einer Rolle in S-Typ bewirkt wird, ist nicht im Störungsgraph-Modul von S-Typ gespeichert, sondern in einem Störungsgraph-Modul eines Einheits-Typs, dem die Besetzung der Rolle angehört.

Faults in an S-type system type can be related to other S-type faults or to S-type roles. You can:

• o As a result of a disturbance that is caused directly from the outside to the S type, is either a disturbance that affects the S type internally, or one that is caused directly from the outside on a role in the S type, or one that is immediate from the S type to the outside.
• o As a result of a disturbance that affects the S type internally, it is either another fault that affects the S type internally, or one that is caused directly from outside on a role in the S type, or one that is caused directly by the S- Type works outwards.
• o The consequence of a fault that has an immediate effect from the S type is not stored in the fault graph module of the S type, but in a fault graph module of a system type in which a system of the type S type one Role occupied.
• o As a result of a disturbance that affects the outside of a role in S-type, is either a disturbance that affects S-type internally, or one that is caused directly from the outside on another role in S-type, or one that acts directly from the S type to the outside.
• o The consequence of a malfunction, which is caused directly on the outside of a role in S type, is not stored in the fault graph module of S type, but in a fault graph module of a unit type to which the role is assigned .

So that the overall fault graph can be created for the technical system, it must be ensured that the logical outer interfaces of the fault graph module for a role match the logical outer interfaces of the fault graph module for the role. A definition can therefore be found when a disturbance in a role in a system type and a disturbance in the occupation of the role correspond to one another. A definition is presented as an advantageous embodiment of the claim. The definition must ensure that there is at most one corresponding disturbance in the staffing of a role for a role. The fact that there is no correspondence in a cast is permissible and means that the functional disturbance described by the role disturbance cannot occur in the cast.

The knowledge module for a unit type On contains all the logical connections between the results of examinations of On and disturbances. Every disturbance for which the knowledge module for on that it explains a non-target result of an examination on on is declared is itself a fault on on and is therefore stored in the knowledge module for on. Any disturbance for which the knowledge module for On is stored, which is incompatible with a result of an examination on On, is itself a disturbance on On.

The computer Wb-Creator automatically creates a variant-specific knowledge base. He has read access to a memory Konf, in which the current configuration of the variant of the technical system for which a knowledge base is to be created is described. The next paragraph describes which information is stored in Konf.

Konf contains a unique identifier for the technical system for each unit. For each identifier, the unit type to which the unit belongs is saved. If n copies of a unit type E-type occur in the technical system, Conf contains n identifiers for these copies, each of which has a reference to the E-type. Furthermore, for a system SYS in Konf is stored, by which unit each role is occupied; SYS can be the technical system as a whole or a subsystem. For each role Ro in SYS, the relationship between SYS, Ro and the unit Ein, which is occupied in SYS Ro, is stored in Conf - for example, as a triple (SYS, Ro, Ein).

The structure of the knowledge modules in the memory WM just shown and the information in the memory Konf is the subject of claim 5.

The procedure according to which the computer Wb creator generates the knowledge base for a variant TS of the technical system is shown below and in FIG. 6. Wb-Creator has read access to the WM memory for the knowledge modules and the Konf memory for the current configuration of the variant. Wb creator has write access to the memory Wb and read and write access to two working memories AS and AS-Stg.

• o Für jede im Wissens-Modul abgespeicherte Störung an E-Typ erzeugt Wb-Ersteller eine Kopie und versieht die Kopie mit einem Verweis auf Ein.
• o Für jeden kausalen Zusammenhang zwischen zwei Störungen, der im Wissens-Modul abgespeichert ist, wird ein Zusammenhang der gleichen Art zwischen den jeweiligen Kopien erzeugt. Falls also im Wissens-Modul für Ein abgespeichert ist, daß St-A Folge von St-B ist, wobei St-A und St-B Störungen an E-Typ sind, so wird im Störungsgraph-Modul für Ein festgehalten, daß die Kopie von St-A mit Verweis auf Ein Folge der Kopie von St-B mit Verweis auf Ein ist.

After read access to WM, Wb-Creator creates a copy of the fault graph module for the technical system as a whole. Wb-Creator takes the information from the storage Konf which units are available in the technical system. For each existing unit On, the Wb creator determines what type the unit belongs to and creates a copy of the fault graph module for the unit type E type of On. The copy is abbreviated to fault graph module for on. To put it more precisely, this "copying" means:

• o For each disturbance to the E-type stored in the knowledge module, Wb-Creator creates a copy and provides the copy with a reference to On.
• o For each causal connection between two disturbances, which is stored in the knowledge module, a connection of the same kind is created between the respective copies. If, therefore, it is stored in the knowledge module for on that St-A is a sequence of St-B, where St-A and St-B are faults on E type, then the fault graph module for on states that the copy from St-A with reference to A sequence of the copy of St-B with reference to On is.

If there are n copies of a unit type in the variant of the technical system, then n copies of the fault graph module are created for the unit type.

All copies of fault graph modules are saved in AS.

• (A) Wb-Ersteller entnimmt dem Speicher Konf die Information, welche Teilsysteme im technischen System vorhanden sind.
• (B) Für jedes SYS, wobei SYS TS selbst oder ein Teilsystem in TS ist,
  • o ermittelt Wb-Ersteller durch Lesezugriff auf Konf, welche Rollen es in SYS gibt, und ermittelt für jede Rolle, durch welche Einheit die Rolle in SYS besetzt ist.
  • o fügt Wb-Ersteller das Störungsgraph-Modul für Ein in das Störungsgraph-Modul für SYS an den Platz von Ro ein.

Once this work is done, the Wb creator creates the overall fault graph for the variant:

• (A) Wb-Creator takes the information from the Konf memory which subsystems are available in the technical system.
• (B) For each SYS, where SYS is TS itself or a subsystem in TS,
  • o Wb creator uses read access to Konf to determine which roles are available in SYS and for each role determines which unit the role in SYS is occupied by.
  • o Wb-Creator inserts the fault graph module for On into the fault graph module for SYS in place of Ro.

• (1) Für jede Störung St im Störungsgraph-Modul für SYS, die von Ro unmittelbar nach außen wirkt, wird im Störungsgraph-Modul für Ein nach einer Störung St-1 gesucht, die von Ein unmittelbar nach außen wirkt und die eine Entsprechung von St ist. Gemäß der Definition von "entsprechend" gibt es maximal eine Entsprechung.
  • Falls Wb-Ersteller eine Entsprechung gefunden hat, wird im Störungsgraph-Modul für SYS St durch St-1 ersetzt. St-1 steht mindestens mit allen Störungen in kausalem Zusammenhang, mit denen St in kausalem Zusammenhang stand.
  • Falls Wb-Ersteller keine Entsprechung gefunden hat, so wird St in den Arbeitsspeicher AS-Stg eingetragen.
• (2) Für jede Störung St im Störungsgraph-Modul für SYS, die unmittelbar von außen an Ro bewirkt wird, wird im Störungsgraph-Modul für Ein nach einer Störung St-1 gesucht, die unmittelbar von außen an Ein bewirkt wird und die eine Entsprechung von St ist. Gemäß der Definition von "entsprechend" gibt es maximal eine Entsprechung.
  • Falls Wb-Ersteller eine Entsprechung gefunden hat, wird im Störungsgraph-Modul für SYS St durch St-1 ersetzt. St-1 steht mindestens mit allen Störungen in kausalem Zusammenhang, mit denen St in kausalem Zusammenhang stand.
  • Falls Wb-Ersteller keine Entsprechung gefunden hat, so wird St in den Arbeitsspeicher AS-Stg eingetragen.

Such an insertion is shown by FIG. 7 and illustrated in FIG. 8, it consists of the following steps:

• (1) For each disturbance St in the disturbance graph module for SYS, which acts directly outwards from Ro, the disturbance graph module for On searches for a disturbance St-1 which acts directly onwards from On and which corresponds to St is. According to the definition of "corresponding", there is at most one correspondence.
  • If Wb creator has found a match, St in the fault graph module for SYS is replaced by St-1. St-1 is at least causally related to all disorders with which St was causally related.
  • If the Wb creator has not found a match, St is entered in the working memory AS-Stg.
• (2) For each fault St in the fault graph module for SYS, which is caused directly from the outside on Ro, the fault graph module for On is searched for a fault St-1, which is caused directly from the outside on and which has a counterpart from St is. According to the definition of "corresponding", there is at most one correspondence.
  • If Wb creator has found a match, St in the fault graph module for SYS is replaced by St-1. St-1 is at least causally related to all disorders with which St was causally related.
  • If the Wb creator has not found a match, St is entered in the working memory AS-Stg.

After all the insertion steps have been completed, the fault graph modules that were previously stored in AS are linked to form a single fault graph. Wb-Creator now removes from this fault graph any disturbance that is also stored in the AS-Stg. A malfunction is stored in the AS-Stg if it is assigned to a role and the Wb creator has not found a corresponding malfunction in the role. Wb-Creator reads in all faults from Wb-Stg and deletes each fault read in itself and iteratively all faults that are direct or indirect consequence of Stg, but not another fault, as well as all faults that directly or indirectly cause Stg, but none other disturbance, from the disturbance graph.

After this work, the fault graph for the variant has been created; Wb creator stores it in the memory Wb.

• o Für jede im Wissens-Modul abgespeicherte Abhilfe, Untersuchung und Resultat erzeugt Wb-Ersteller eine Kopie und versieht die Kopie mit einem Verweis auf Ein.
• o Für jeden Zusammenhang zwischen einer Einfach-Ursache und einer Abhilfe, einer Untersuchung und einem Resultat und zwischen einer Störung und einem Resultat, der im Wissens-Modul für E-Typ abgespeichert ist, wird ein Zusammenhang der gleichen Art zwischen den jeweiligen Kopien erzeugt.

Wb creator again determines by read access to the memory Konf which units occur in the technical system for which a knowledge base is to be created. For each unit On, where On is the technical system as a whole or a unit in the technical system, Wb-Konf determines which type E type Ein belongs to and copies the knowledge module for E type in addition to the previously copied fault graph -Module. To put it more precisely, this "copying" means:

• o For each remedy, examination and result stored in the knowledge module, Wb-Creator creates a copy and provides the copy with a reference to On.
• o For every connection between a simple cause and a remedy, an examination and a result and between a fault and a result that is stored in the knowledge module for E-Type, a connection of the same kind is created between the respective copies.

Wb-Creator stores all copies of the knowledge modules created in this way in Wb. The knowledge base is now configured.

The device according to the invention solves the task set at the beginning of the description of efficiently creating knowledge bases for each variant of a class of technical systems. One has to create a knowledge module for each unit type. For each variant for which a knowledge base is to be created, the current configuration must be described according to the invention. A knowledge base configuration system, which has read access to all knowledge modules and all current configurations, then creates all the necessary knowledge bases fully automatically.

If in a variant a unit is replaced by a different type of unit, the new configuration of the variant is described according to the invention. Then the knowledge base configuration system automatically creates a complete new knowledge base for the variant.

The device according to the invention is suitable for automatically generating a fault model from fault graph modules and the current configuration of TS, namely an overall fault graph for TS (claim 4). To solve this task, a knowledge module only needs to contain the fault graph module for the respective unit type. The process initially works like the one used to create a knowledge base for an expert system. However, it has already achieved its goal and stops when the overall fault graph for TS is completely generated.

Advantageous design

The device according to the invention provides a fault graph and all other information so that an expert system can use the troubleshooting tree method.

This path, which is in principle feasible, has a serious disadvantage: in order for an expert system to be able to get from the node it has currently reached to a leaf, it must change from the node to one of its successors. It is often not possible to clearly identify one of the successors based on the results obtained. Then the expert system has to "generate a hypothesis", that is, to continue testing in one of the successors on a trial basis. If the hypothesis later turns out to be wrong, the expert system has to return to the node, ie carry out a so-called "backtracking", which can be time-consuming and lead to inconsistencies.

• o Eine Untersuchung, deren Resultate zwischen den Nachfolgern zu differenzieren vermag, gibt es nicht.
• o Da eine Mehrfach-Ursache vorliegt, werden mehrere Nachfolger identifiziert.

The following reasons can be responsible for the expert system not being able to clearly identify a successor to the node reached:

• o There is no investigation whose results can differentiate between the successors.
• o Since there is a multiple cause, several successors are identified.

Another way is therefore taken in sub-claim 6. The expert system does not "navigate" in the fault graph with the aim of reaching a page. Rather, the expert system diagnoses the technical system using a method which is illustrated in FIG. 9 and which can find not only a single but also a multiple cause by means of a single diagnosis.

During the diagnosis, the expert system has read and write access to a working memory AS-declaration, in which all possible causes are still stored. "Still possible" is a cause if it explains every non-target result obtained in the course of the diagnosis so far and is incompatible with none. A multiple cause explains a result if at least one of the single causes that make up the multiple cause explains the result. It is incompatible with a result if at least one of its simple causes is incompatible with the result. A multiple cause is considered if it is still possible, but none of its individual causes is still possible.

The expert system enters any causes in a second working memory AS-non that are incompatible with a result obtained.

• o Zu Beginn der Diagnose sind noch alle Ursachen möglich. In den Arbeitsspeicher AS trägt das Expertensystem daher jedes Blatt des Störungsgraphen ein.
• o Das Expertensystem wiederholt die folgende Abfolge von Schritten so lange, bis im Arbeitsspeicher AS-erkl nur noch eine einzige Ursache enthalten ist:
  • (1) Das Expertensystem bestimmt, welche Untersuchung als nächstes durchgeführt wird
  • (2) Das Expertensystem selber oder ein Mensch führt diese Untersuchung durch und ermittelt, welche der möglichen Resultate die Untersuchung aktuell hat.
  • (3) Das Expertensystem wertet das in Schritt (2) gewonnene Resultat aus und ändert den Inhalt von AS-erkl ab. Wie dies geschieht, wird im nächsten Absatz beschrieben.
  • (4) Das Expertensystem prüft, wie viele Ursachen sich jetzt in AS-erkl befinden.
• o Sobald in AS nur noch eine einzige Ursache enthalten ist, ist diese Ursache als tatsächlich vorliegend identifiziert. Für jede ihrer Einfach-Ursachen wird die zugeordnete Abhilfe entweder automatisch vom Expertensystem ausgeführt oder einem Menschen zur Ausführung vorgeschlagen.

The expert system proceeds with the diagnosis as follows:

• o At the beginning of the diagnosis, all causes are still possible. The expert system therefore enters each sheet of the fault graph in the working memory AS.
• o The expert system repeats the following sequence of steps until there is only one cause in the working memory AS-explain:
  • (1) The expert system determines which examination is to be carried out next
  • (2) The expert system itself or a person carries out this examination and determines which of the possible results the examination currently has.
  • (3) The expert system evaluates the result obtained in step (2) and changes the content of AS-explain. How this happens is described in the next paragraph.
  • (4) The expert system checks how many causes are now in AS-explain.
• o As soon as there is only one cause in AS, this cause is identified as actually present. For each of its simple causes, the assigned remedy is either carried out automatically by the expert system or suggested to a person for execution.

10 and 11 illustrate the work steps by which a result Resu is evaluated and the evaluation result is written into the working memory AS-explain:

• o Das Expertensystem ermittelt, welche Ursachen mit Resu unvereinbar sind.
• o Das Expertensystem streicht aus AS-erkl alle Ursachen, die mit Resu unvereinbar sind.
• o Das Expertensystem stellt fest, ob Resu das Soll-Resultat oder ein Nicht-Soll-Resultat derjenigen Untersuchung ist, die Resu lieferte
• o Falls es ein Nicht-Soll-Resultat ist, so macht das Expertensystem folgendes:
  • Es ermittelt, welche Ursachen Resu erklären.
  • Es streicht aus AS-erkl alle Ursachen, die Resu nicht erklären.
  • Es ergänzt AS-erkl um jede mögliche Mehrfach-Ursache, die
    • aus einer Ursache aus AS-erkl, die Resu nicht erklärt, aber auch nicht mit Resu unvereinbar ist
    • und einer Ursache, die Resu erklärt und die weder in AS-erkl noch in AS-unver enthalten ist, d. h. die zwar nicht alle bisher gewonnenn Resultate erklärt, aber mit keinem unvereinbar ist
    besteht.
• o Das Expertensystem ergänzt AS-unver um alle Ursachen, die mit Resu unvereinbar sind.

The Resu result is evaluated as follows:

• o The expert system determines which causes are incompatible with Resu.
• o The expert system removes all causes that are incompatible with Resu.
• o The expert system determines whether Resu is the target result or a non-target result of the examination that Resu delivered
• o If it is a non-target result, the expert system does the following:
  • It determines what causes Resu to explain.
  • It omits AS-explain all causes that Resu does not explain.
  • It complements AS-explain with every possible multiple cause
    • for a cause from AS-explain that Resu does not explain but is also not incompatible with Resu
    • and a cause which Resu explains and which is neither contained in AS-explain nor in AS-non, ie which does not explain all the results obtained so far, but which is incompatible with none
    consists.
• o The expert system supplements AS-unver with all causes that are incompatible with Resu.

• (1) Falls eine Störung St ein Resultat Resu erklärt, so erklärt auch jeder Verursacher von St Resu.
• (2) Falls eine Störung St mit einem Resultat Resu unvereinbar ist, so ist ein Verursacher St-1 von St

In order for the expert system to be able to evaluate a resultant resu, the information must be available, which causes explain it and which relationships are incompatible with it. The expert system obtains this information by combining the relationships between resu and faults stored in the knowledge base with the overall fault graph for the technical system. It applies these principles:

• (1) If a disturbance St explains a result of Resu, every cause of St Resu also explains.
• (2) If a disturbance St is incompatible with a result Resu, a cause St-1 of St

• (A) Ermittelt wird, welche Störungen in der Wissensbasis als Erklärungen von Resu abgespeichert sind.
• (B) Für jede Erklärung St-E, die selbst nicht Ursache ist, werden die Ursachen ermittelt, die St-E verursachen: Es sind alle Blätter desjenigen Astes des Störungsgraphen, dessen Wurzel St-E ist.
• (C) Jede Ursache, die entweder direkt in der Wissensbasis als Erklärung von Resu abgespeichert ist oder die in Schritt (B) mindestens einmal gefunden wurde, erklärt Resu.

Principle (1) is used as follows to determine what causes Resu:

• (A) It is determined which faults are stored in the knowledge base as explanations from Resu.
• (B) For every explanation St-E that is not itself the cause, the causes that cause St-E are determined: all the leaves of the branch of the fault graph whose root is St-E.
• (C) Resu explains any cause that is either stored directly in the knowledge base as an explanation of Resu or that was found at least once in step (B).

• (A) Ermittelt wird, für welche Störungen in der Wissensbasis abgespeichert ist, daß sie mit Resu unvereinbar sind.
• (B) Für jede mit Resu unvereinbare Störung St-E, die selbst nicht Ursache ist, werden die Ursachen ermittelt, die St-E verursachen: Es sind alle Blätter desjenigen Astes des Störungsgraphen, dessen Wurzel St-E ist.
• (C) Eine Ursache Urs, die in (C) als unvereinbar ermittelt wurde, ist dann mit Resu unvereinbar, wenn jede Folge von Urs entweder selbst mit Resu unvereinbar ist oder Verursacher einer mit Resu unvereinbaren Störung ist.
• (D) Außerdem ist jede Ursache mit Resu unvereinbar. die direkt in der Wissensbasis als mit Resu unvereinbar abgespeichert ist.

Principle (2) is used as follows to determine what causes are inconsistent with Resu:

• (A) It is determined for which faults are stored in the knowledge base that they are incompatible with Resu.
• (B) For every fault St-E incompatible with Resu that is not itself the cause, the causes that cause St-E are determined: all the leaves of the branch of the fault graph whose root is St-E.
• (C) A cause Urs, which was determined in (C) as incompatible, is incompatible with Resu if each episode of Urs is either incompatible with Resu itself or is the cause of a fault incompatible with Resu.
• (D) In addition, every cause is incompatible with Resu. which is stored directly in the knowledge base as incompatible with Resu.

• o abgespeichert, welche Einfach-Ursachen an TS und seinen Modulen auftreten können,
• o für jede dieser Ursachen abgespeichert, durch welche Abhilfe die Ursache beseitigt wird,
• o abgespeichert, welche Untersuchungen an TS und seinen Modulen im Verlaufe einer Diagnose durchgeführt werden können,
• o für jede dieser Untersuchungen abgespeichert, welches Soll-Resultat und welche Nicht-Soll-Resultate die Untersuchung hat,
• o für jedes Resultat abgespeichert, mit welchen Ursachen es unvereinbar ist,
• o für jedes Nicht-Soll-Resultat zusätzlich abgespeichert, durch welche Ursachen es erklärt wird.

An advantageous embodiment of the invention is that a computer Ex in a first phase "extracts" from the automatically generated knowledge base for the technical system TS to be diagnosed the information that is required for the diagnostic strategy according to the invention, and in a memory Wb-ex discards. After the work of Ex is in Wb-ex

• o which simple causes can occur on TS and its modules,
• o stored for each of these causes, by which remedy the cause is eliminated,
• o stored which examinations can be carried out on TS and its modules in the course of a diagnosis,
• o for each of these examinations, which target result and which non-target results the examination has, are stored,
• o stored for each result, with what causes it is incompatible,
• o For each non-target result, additionally saved, by what causes it is explained.

In a second phase, the expert system works according to the diagnostic strategy according to the invention and has read access to the memory Wb-ex. The expert system does not need the "actual" knowledge base that the computer Wb created and stored in the memory Wb. This procedure is advantageous because the extraction only needs to be carried out once and the result of the extraction can be reused for every diagnosis by TS as long as no diagnosis-relevant change is made to TS. In addition, the extraction result requires less storage space than the original knowledge base.

How should the expert system determine which examination is to be carried out next during a diagnosis ?: One possibility is to ask the user. However, this may overwhelm inexperienced users. It is advantageous that when the user does not want or cannot make a selection, the expert system automatically determines the next examination in each case. A suitable selection strategy significantly shortens the duration of the diagnosis.

An advantageous embodiment of the invention, claim 8 is: The expert system selects the examination that has the greatest relationship between the benefits and the costs. For this purpose, the expert system calculates a numerical evaluation for the benefit N (Unt) for the possible causes and the costs K (Unt) for each examination still under consideration, and then puts N (Unt) and K (Unt) in relation. The examination with the greatest ratio is selected.

wobei für i=1,...,n q-i die bedingte Wahrscheinlichkeit ist, daß Unt das Resultat Resu-i hat, falls alle Ursachen in UM und keine anderen Ursachen noch möglich sind, und
IG-i = - Id[P(UM-i)/P(UM)] der Informationsgewinn durch das Resultat Resu-i ist.The usefulness of an investigation Unt is assessed as the expected information gain according to a method that is known from information theory. Let UM be the set of possible causes before performing Unt. Let Resu-1, ..., Resu-n be the possible results of Unt, and for i = 1, ..., n UM-i be the set of causes that are still possible if the result Resu-i is won. Then $$\text{N (Unt) = q-1 * IG-1 + ... + qn * IG-n,}$$

where for i = 1, ..., n qi is the conditional probability that Unt has the result Resu-i if all causes in UM and no other causes are still possible, and
IG-i = - Id [P (UM-i) / P (UM)] is the information gain from the result Resu-i.

In order to determine N (Unt) for an examination Unt, a characterization of the frequency of occurrence of each individual cause is required in order to calculate the probability of occurrence of every possible cause. In order to determine K (Unt), a characterization of the costs that the implementation of Unt causes is needed.

In order to achieve a systematic structuring of information about unit types and to reduce the number of entries, similar unit types are grouped into superclasses. All common information is noted in the superclasses, while in the unit type itself the individual information including the affiliation to superclasses is entered. The knowledge module for a unit type A includes information about which superclasses A has. This procedure is known as the taxonomic organization of the diagnostic information, because a taxonomy (kinship hierarchy) is established under the unit types. This taxonomy forms a directed graph with roots, whose nodes stand for unit types and whose edges stand for the relation "has the upper classes". The root stands for "unit type", the leaves for actually used unit types.

FIG. 13 explains how information about a unit type E type is accessed. The basic procedure is to recursively "collect" the required information from the information about the E type and the superclasses of the E type.

As an example, we are looking for information on which roles in E-type must be fulfilled so that E-type can fulfill its intended function. To do this, the E-Type knowledge module first determines which roles are stored there in E-Type, and these roles are buffered. Then determined which upper classes have E-type. For each superclass E-type-whether, the knowledge module for E-type-whether determines which roles are stored there in E-type-whether, and the buffer for roles is supplemented. This procedure is continued for the upper classes of E-type-ob, it is finished when the root of the taxonomy tree is reached.

The advantageous embodiment of the invention, to which claim 14 applies, solves the problem of how it is ensured at the time of diagnosis that the expert system has access to the memory Konf for the current configuration of the technical system to be diagnosed at the time of diagnosis and that this memory with the actual configuration matches. This problem occurs e.g. B. then when the expert system works in a car workshop.

The advantageous embodiment of the invention provides that the memory Konf is contained in the technical system to be diagnosed itself. When a copy of a technical system is delivered, the memory Konf is filled with the information about its configuration structured according to the invention. Whenever a module of the item is replaced by a different one, the content of the memory Konf is updated.

A further requirement is dealt with by the advantageous embodiment of the invention to which claim 14 applies: it must be ensured that the expert system at the time of diagnosis has read access to each knowledge module for a unit type which is used in the technical system to be diagnosed. For example, in the case of the expert system in the vehicle workshop, this can lead to logistical problems (in the "delivery" of the knowledge modules) and storage space problems, especially when technical, when all knowledge modules are stored in the vehicle workshop itself Changes to unit types often force new versions of knowledge modules to be provided. If, on the other hand, each knowledge module is saved only once, a lot of data must be made available to the expert system quickly at the time of diagnosis.

The advantageous embodiment of the invention (claim 6) provides that all the necessary knowledge modules are stored in the technical system itself. One way of realizing this idea is that the knowledge module for the type of one is stored in each unit one. If one unit is replaced by another, one only has to ensure that the new unit also contains "its" knowledge module, in order to ensure that all knowledge modules are available to the expert system even after the exchange. Storage space is saved if space-consuming information such as B. Text in the technical system can be stored coded. One must then ensure that the expert system has the necessary information to decode the coding, e.g. B. by the expert system has read access to a table through which text elements and associated code are linked.

For example, in computer-aided diagnosis in the automotive workshop, an expert system must be able to diagnose all or at least many variants of a technical system. It can take significant time and large logistical and storage space problems, e.g. B. in the case of the expert system in the vehicle workshop, if the knowledge base for each variant to be diagnosed is created centrally in advance and delivered to all locations; Large data can be transferred quickly if each knowledge base is only kept centrally and the expert system must have read access to the centrally stored knowledge bases in the event of a diagnosis.

Therefore, if an expert system must be able to diagnose several variants, it is advantageous that the knowledge base for the variant currently to be diagnosed is only generated at the time of diagnosis (claim 12). In order to avoid time-consuming data transfers, the same software and hardware system expediently first configures the knowledge base for the variant and then carries out the diagnosis with this knowledge base.

• o welche Einheiten in TS vorhanden sind,
• o von welchem Typ TS ist,
• o für jede Einheit Ein in TS, von welchem Typ Ein ist, und
• o für jedes SYS, wobei SYS TS selber oder ein System in TS ist, und für jede Rolle in SYS die Kennzeichnung, welche Einheit in SYS diese Rolle besetzt.

The device according to the invention requires a memory Konf which contains the following information about the variant TS of the technical system for which a knowledge base is to be created:

• o which units are in TS,
• o which type is TS,
• o for each unit on in TS, of which type is on, and
• o for each SYS, where SYS is TS itself or a system in TS, and for each role in SYS the identification which unit in SYS occupies this role.

In principle, it is possible to generate this information "by hand". But they are characteristic of the TS variant and cannot be reused for other variants. In addition, similarities between different subsystems of the same type cannot be systematically exploited in the variant. It is therefore advantageous if the content of Konf is generated automatically from information which is assigned to the unit types and not to the unit instances.

The following describes what configuration information is assigned to each unit type in addition to its knowledge module and how the description according to the invention of the current configuration of the technical system TS, for which a knowledge base is to be created, is automatically generated from this configuration information .

• (1) Für jede Rolle Ro in S-Typ ist festgelegt, welchem Einheits-Typ Ein die Besetzung von Ro angehört. Zwei Arten von Festlegungen werden unterschieden:
  • Entweder ist Ro in S-Typ eine einzige unbedingte Festlegung zugeordnet, nämlich die, daß in jedem Exemplar vom Typ S-Typ die Rolle Ro durch eine Einheit vom Typ Ein besetzt ist,
  • oder Ro in S-Typ sind mehrere bedingte Festlegungen zugeordnet. Eine bedingte Festlegung nennt einen Einheits-Typ Ein und als Bedingung eine Rolle Ro-1 in einem System-Typ S1-Typ. Die Festlegung bedeutet, daß in einem Exemplar Sys-Ex vom Typ S-Typ die Rolle Ro dann durch eine Einheit vom Typ Ein besetzt wird, wenn Sys-Ex die Rolle Ro-1 in einem System vom Typ S1-Typ besetzt.
• (2) Für eine Rolle Ro in S-Typ kann festgelegt sein, welche anderen Rollen in einem Exemplar vom Typ S-Typ eine Besetzung von Ro ebenfalls besetzt. Der E-Motor einer Straßenbahn ist ein Beispiel für eine Einheit, die sowohl die Rolle "Antrieb" als auch die Rolle "Induktionsbremse" im technischen System "Straßenbahn" besetzt. Auch hier werden zwei Arten von Festlegungen unterschieden:
  • Eine unbedingte Festlegung nennt ausschließlich eine andere Rolle Ro-1 in S-Typ. Diese Festlegung bedeutet, daß in jedem Exemplar vom Typ S-Typ diejenige Einheit, die die Rolle Ro besetzt, zugleich auch Ro-1 besetzt.
  • Eine bedingte Festlegung nennt eine andere Rolle Ro-1 in S-Typ und als Bedingung eine Rolle Ro-2 in einem anderen System-Typen S2-Typ. Die Festlegung bedeutet, daß in einem Exemplar Sys-Ex des Typs S-Typ die Besetzung von Ro dann zugleich Ro-1 besetzt, wenn Sys-Ex die Rolle Ro-2 in einem System vom Typ S2-Typ besetzt.
• (3) Für eine Rolle Ro in S-Typ kann festgelegt sein, welche Rollen in anderen Systemen eine Besetzung von Ro in einem Exemplar vom Typ S-Typ ebenfalls besetzt. Zwei Arten von Festlegungen werden unterschieden:
  • Eine unbedingte Festlegung nennt ausschließlich eine Rolle Ro-1 in einem anderen System-Typen S1-Typ. Diese Festlegung bedeutet, daß diejenige Einheit Ein, die in jedem Exemplar vom Typ S-Typ die Rolle Ro besetzt, zugleich auch in jedem Exemplar vom Typ S1-Typ die Rolle Ro-1 besetzt. Die Festlegung darf nur dann getroffen werden, wenn es nur eine einzige derartige Einheit Ein gibt.
  • Eine bedingte Festlegung nennt eine Rolle Ro-1 in einem anderen System-Typen S1-Typ und
    • entweder nur eine Bedingung für S-Typ, bestehend aus einer Rolle Ro-2 in einem von S-Typ verschiedenen System-Typen S2-Typ (Bedingung vom Typ I)
    • oder nur eine Bedingung für S1-Typ, bestehend aus einer Rolle Ro-2 in einem von S1-Typ verschiedenen System-Typen S2-Typ (Bedingung vom Typ II)
    • oder je eine Bedingung für S-Typ und eine für S1-Typ, wobei die erste Bedingung aus einer Rolle Ro-2 in einem von S-Typ verschiedenen System-Typen S2-Typ und die zweite Bedingung aus einer Rolle Ro-3 in einem von S1-Typ verschiedenen System-Typen S3-Typ besteht (Bedingung vom Typ III).

Let S-type be a system type that occurs in TS; S-type is either the type to which the technical system belongs, or a type of a subsystem used in TS. The following information is assigned to the S type:

• (1) For each role Ro in S-type it is determined which unit type On the cast of Ro belongs to. There are two types of definitions:
  • Either Ro is assigned a single unconditional definition in S-type, namely that in each S-type copy the role Ro is occupied by a unit of type On,
  • or Ro in S-type are assigned several conditional definitions. A conditional definition calls a unit type On and as a condition a role Ro-1 in a system type S1 type. The definition means that in a Sys-Ex of type S type, the role Ro is occupied by a unit of type On if Sys-Ex occupies the role Ro-1 in a system of type S1.
• (2) For a role Ro in S-type it can be determined which other roles in a copy of the type S-type also occupy Ro. The electric motor of a tram is an example of a unit that occupies both the role "drive" and the role "induction brake" in the technical system "tram". There are also two types of definitions:
  • An unconditional definition only calls another role Ro-1 in S-type. This definition means that in each S-type copy the unit that occupies the role Ro also occupies Ro-1.
  • A conditional definition specifies another role Ro-1 in S type and as a condition a role Ro-2 in another system type S2 type. The definition means that in a copy of the type S-type Sys-Ex, the occupation of Ro also occupies Ro-1 if Sys-Ex occupies the role Ro-2 in a system of the type S2-type.
• (3) For a role Ro in S-type it can be determined which roles in other systems also occupy Ro in a copy of type S-type. There are two types of definitions:
  • An unconditional definition only refers to a role Ro-1 in another system type S1 type. This definition means that the unit Ein which occupies the role Ro in every copy of the type S, also simultaneously occupies the role Ro-1 in every copy of the type S1. The determination may only be made if there is only one such unit, On.
  • A conditional definition calls a role Ro-1 in another system type S1 type and
    • either only one condition for S-type, consisting of a role Ro-2 in a system type that is different from S-type S2-type (condition of type I)
    • or just a condition for S1 type, consisting of a role Ro-2 in a system type different from S1 type S2 type (condition of type II)
    • or one condition each for S type and one for S1 type, the first condition consisting of a role Ro-2 in a system type different from S type S2 type and the second condition consisting of a role Ro-3 in one system type different from S1 type exists (condition of type III).

• Sys-ex in einem Exemplar vom Typ S2-Typ die Rolle Ro-2 besetzt
• bzw. Sys-1-ex in einem Exemplar vom Typ S2-Typ die Rolle Ro-2 besetzt
• bzw. sowohl Sys-ex in einem Exemplar vom Typ S2-Typ die Rolle Ro-2 besetzt als auch Sys-1-ex in einem Exemplar vom Typ S2-Typ die Rolle Ro-2 besetzt.

This conditional stipulation means that the unit Ein that occupies the role Ro in an instance of Sys-ex of type S type also always occupies role Ro-1 in an instance of Sys-1-ex of type S1, if

• Sys-ex in a copy of the type S2-type occupies the role Ro-2
• or Sys-1-ex in a copy of the type S2-type occupied the role Ro-2
• or both Sys-ex in a copy of type S2 occupy the role Ro-2 and Sys-1-ex in a copy of type S2 occupy the role Ro-2.

The configuration information is assigned to unit types, just like the knowledge modules structured according to the invention. However, the configuration information about a system type S type - in contrast to that in the knowledge module for S type - depends on the use of a copy of Sys-Ex of type S type and on the version of the subunits in Sys -Ex. For this reason, they are expediently kept in a separate memory conf information. The structuring of the information in conf info applies to claim 9.

The description of the current configuration of a technical system TS constructed according to the invention is generated by a computer Konf-R and stored in a memory Konf. Konf-R has read access to the memory WM structured according to the invention with the knowledge modules, to the memory Konf-Info structured according to the advantageous embodiment, write access to the memory Konf, which contains the current configuration after the work of Konf-R, and read - and write access to a working memory AS.

In order for the Konf-R computer to work, a method must be developed and used by Konf-R in order to automatically generate a unique identifier for each unit in the technical system. This identifier expediently depends on the type of which one is, and - unless one is the technical system itself - on the identifier of a system Sys-ex in which one has a role and on the identifier of this role.

Fig. 12 explains the procedure of Konf-R for a technical system TS:

• o Konf-R ermittelt, welchem Typ TS-Typ TS angehört, und speichert diese Information in Konf ab
• o Konf-R ermittelt durch Lesezugriff auf WM, welche Rollen in TS-Typ existieren.
• o Für jede derartige Rolle Ro trägt Konf-R in AS ein Tupel ein, bestehend aus der Kennung für TS und der für Ro.
• o Konf-R durchläuft die folgende Schleife so oft, bis AS leer ist:
  • Konf-R ermittelt in AS jedes Tupel, das ein System SYS und eine Rolle Ro in SYS bezeichnet, für dessen Besetzung Ein die Kennung bereits erzeugt wurde. Daß die kennung für Ein bereits erzeugt wurde, ist dann möglich, wenn Ein zugleich eine andere Rolle besetzt. Wie Konf-R diese Tupel ermittelt, schildert der nächste Absatz.
  • Für jedes derartige Tupel
    • schreibt Konf-R ein Tripel in Konf, das aus der Kennung von SYS, der von Ro und der von Ein besteht.
    • löscht Konf-R in AS das Tupel
  • Falls AS noch Tupel enthält, so vollführt Konf-R folgende Arbeitsschritte:
    • Konf-R wählt ein anderes Tupel aus AS aus. Sei SYS das durch das Tupel bezeichnete System und Ro die bezeichnete Rolle in SYS.
    • Konf-R stellt durch Lesezugriff auf Konf fest, von welchem Typ SYS ist
    • Konf-R stellt durch Lesezugriff auf Konf-lnfo fest, von welchem Typ E-Typ die Besetzung Ein von Ro in SYS ist.
    • Konf-R erzeugt Konf-R eine Kennung für Ein, die von E-Typ, Ro und SYS abhängt.
    • Konf-R trägt in Konf ein Tripel ein, bestehend aus den Kennungen für SYS, Ro und Ein
    • Konf-R trägt in Konf ein, daß Ein vom Typ E-Typ ist, d. h. trägt in Konf einen Verweis von der Kennung für Ein auf die für E-Typ ein.
    • Konf-R löscht in AS das Tupel
    • Konf-R ermittelt durch Lesezugriff auf WM, ob E-Typ ein System- oder ein Komponenten-Typ ist
    • Falls E-Typ ein System-Typ ist,
      • ermittelt Konf-R alle Rollen in E-Typ
      • und trägt für jede Rolle in E-Typ eine Kennung in AS ein, die aus einer Kennung für Ein und einer für die Rolle besteht.
• o Falls AS leer ist, so ist die aktuelle Konfiguration vollständig und erfindungsgemäß in Konf abgespeichert.

Each entry in the working memory AS is a tuple, consisting of an identifier for TS or a subsystem SYS from TS and an identifier for a role in TS or SYS. For each pair identified by a tuple in AS, the occupation must still be determined. The entries in Konf are generated by Konf-R, each entry is a triple, consisting of an identifier for TS or a subsystem SYS from TS, an identifier for a role Ro in TS or SYS and an identifier for the unit Ein, die Ro in TS or SYS occupied.

• o Konf-R determines which type TS type TS belongs to and stores this information in Konf
• o Konf-R uses read access to WM to determine which roles exist in the TS type.
• o For each such roll Ro, Konf-R enters a tuple in AS, consisting of the identifier for TS and that for Ro.
• o Conf-R runs through the following loop until AS is empty:
  • Konf-R determines in AS every tuple that designates a system SYS and a role Ro in SYS, for the occupation of which the identifier has already been generated. It is possible that the identifier for one has already been generated if one also has a different role. The next paragraph describes how Konf-R determines these tuples.
  • For each such tuple
    • Konf-R writes a triple in Konf, which consists of the identifier of SYS, that of Ro and that of Ein.
    • Konf-R deletes the tuple in AS
  • If AS still contains tuples, Konf-R carries out the following work steps:
    • Konf-R selects another tuple from AS. Let SYS be the system designated by the tuple and Ro be the designated role in SYS.
    • Konf-R uses read access to Konf to determine which type SYS is
    • Konf-R uses read access to Conf-Info to determine which type E-type the occupation is Ro by in SYS.
    • Konf-R generates Konf-R an identifier for on, which depends on E type, Ro and SYS.
    • Konf-R enters a triple in Konf, consisting of the identifiers for SYS, Ro and Ein
    • Konf-R enters in Konf that one is of type E-type, ie in Konf enters a reference from the identifier for On to that for E-type.
    • Conf-R deletes the tuple in AS
    • Konf-R uses read access to WM to determine whether the E type is a system or a component type
    • If E-type is a system type,
      • Konf-R determines all roles in E type
      • and enters an identifier in AS for each role in E type, which consists of an identifier for one and one for the role.
• o If AS is empty, the current configuration is complete and, according to the invention, is stored in Konf.

• (A) Konf-R ermittelt den Typen S-Typ von Sys-Ex und stellt durch Lesezugriff auf den Speicher Konf-lnfo fest, welche Festlegungen der Art (2) für die Rolle Ro in S-Typ in Konf-lnfo abgespeichert sind.
  • Für jede Festlegung, die Ro in S-Typ mit einer anderen Rolle Ro-1 in S-Typ verknüpft, sucht Konf-R nach einem Tripel in Konf, das aus der Kennung für Sys-Ex, der für Ro-1 und der für eine Einheit Ein besteht. Ist die Festlegung in Konf-lnfo eine bedingte, stellt Konf-R durch Lesezugriff auf Konf fest, ob Sys-Ex die Bedingung erfüllt. Ist die Bedingung erfüllt oder die Festlegung eine unbedingte, so ist Ein die gesuchte Besetzung, die Kennung von Ein steht in Konf.
  • Findet Konf-R mehrere passende Einheiten Ein, so sind die in Konf-lnfo abgespeicherten Informationen widersprüchlich. Konf-R bricht seine Arbeit ab und meldet den entdeckten Widerspruch.
• (B) Falls Konf-R durch Schritt A keine passende Einheit gefunden hat, so stellt Konf-R durch Lesezugriff auf den Speicher Konf-lnfo fest, welche Festlegungen der Art (3) für die Rolle Ro in S-Typ in Konf-lnfo abgespeichert sind.
  • Für jede Festlegung, die Ro in S-Typ mit einer Rolle Ro-1 in einem anderen System-Typ S1-Typ verknüpft, sucht Konf-R in Konf nach allen Kennungen für Exemplare vom Typ S1-Typ. Falls es derartige Kennungen gibt, seien Sys-Ex-1, ..., Sys-Ex-n die durch die Kennungen benannten Exemplare des Typs S1-Typ.
    • Für jedes Sys-Ex-i (i=1,...,n) sucht Konf-R nach einem Tripel in Konf, das aus der Kennung für Sys-Ex-i, der für Ro-1 und der für eine Einheit Ein besteht. Ist die Festlegung eine bedingte der Art (Typ I), prüft Konf-R durch Lesezugriff auf Konf, ob Sys-Ex die Bedingung erfüllt; ist sie eine bedingte der Art (Typ II), prüft Konf-R, ob Sys-Ex-i die Bedingung erfüllt; ist sie eine bedingte der Art (Typ III), prüft Konf-R, ob sowohl Sys-Ex als auch Sys-Ex-i jeweils die Bedingung erfüllen.
  • Findet Konf-R mehrere passende Einheiten Ein, so sind die in Konf-lnfo abgespeicherten Informationen widersprüchlich. Konf-R bricht seine Arbeit ab und meldet den entdeckten Widerspruch.

Through steps (A) and (B) described below, Konf-R determines whether a tuple in the working memory AS designates a system and a role, for the occupation of which an identifier has already been generated: Be Sys-Ex the system and Ro the Role in the system from which the tuple consists of identifiers.

• (A) Konf-R determines the type S-type from Sys-Ex and uses read access to the memory Konf-Info to determine which specifications of type (2) for the role Ro in S-type are stored in the conf-info.
  • For each definition that links Ro in S-type to another role Ro-1 in S-type, Konf-R searches for a triple in Konf that consists of the identifier for Sys-Ex, that for Ro-1 and that for one unit one exists. If the definition in Konf-Info is conditional, Konf-R uses read access to Konf to determine whether Sys-Ex fulfills the condition. If the condition is fulfilled or the specification is unconditional, then On is the desired occupation, the identifier of On is in Conf.
  • If Konf-R finds several suitable units, the information stored in Konf-Info is contradictory. Konf-R stops working and reports the discovered contradiction.
• (B) If Konf-R has not found a suitable unit through step A, Konf-R uses read access to the memory Konf-Info to determine which specifications of type (3) for the role Ro in S-type in Conf-Info are saved.
  • For each definition that links Ro in S type to a role Ro-1 in another system type S1 type, Konf-R in Konf searches for all identifiers for copies of type S1 type. If there are such identifiers, Sys-Ex-1, ..., Sys-Ex-n are the instances of the S1 type designated by the identifiers.
    • For each Sys-Ex-i (i = 1, ..., n) Konf-R searches for a triple in Konf, which consists of the identifier for Sys-Ex-i, that for Ro-1 and that for one unit consists. If the definition is a conditional of the type (type I), Konf-R uses read access to Konf to check whether Sys-Ex fulfills the condition; if it is a conditional of the type (type II), Konf-R checks whether Sys-Ex-i fulfills the condition; if it is a conditional of the type (type III), Konf-R checks whether both Sys-Ex and Sys-Ex-i each meet the condition.
  • If Konf-R finds several suitable units, the information stored in Konf-Info is contradictory. Konf-R stops working and reports the discovered contradiction.

Claims (18)

1. Equipment for compiling the knowledge base of an expert system which diagnoses a technical system consisting of modules,

   consisting of a computer (Wb compiler) with reading access to a first memory (WM) and a second memory (Konf) as well as at least temporarily with writing access to a third memory (Wb),

   from information data about the technical system, about its disturbances and about its diagnosis, which are filed so structured in the first memory (WM), that it gives a knowledge module which contains all needed information data about the internal build-up of a technical system of the type to which the technical system belongs, as well as in case of need about the disturbances, remedies and examinations at a technical system of this type,

   as well as gives a respective knowledge module which contains all needed information data about the internal build-up of a module of each type of module which occurs at least once in the technical system, as well as in case of need about the disturbances, remedies, examinations and results at a module of this type,

   characterised thereby, that

   the information data as to which roles, i.e. which partial functions that are fulfillable by association of a module, must be cast, i.e. be associated with a module, in order that the technical system can fulfil its target function, is contained in the knowledge module for the type to which the technical system belongs,

   the information data as to which roles must be cast in a module of each type of composite modules, which occur in a technical system of that type, in order that the module can fulfil its target function, is contained in the knowledge module for that module type, the information is available for each role in the technical system and each role in each composite module in the technical system as to which module in the technical system or in the composite module is cast in the role,

   the configuration of the technical system is filed in the second memory (Konf) and in the third memory (Wb),

   o there is contained a knowledge module for the technical system as copy of the knowledge module for the type to which the technical system belongs,

   o there is contained an individual knowledge module as copy of the knowledge module for the type, to which the module belongs, for each module of the technical system and

   o each knowledge module formed in such a manner for a module (A) according to the role which module (A) occupies in the composite module (B) or in the technical system is put together with the knowledge module formed in such a manner for the module (B) or for the technical system.
2. Equipment according to claim 1,
   characterised thereby, that

   an unambiguous identification of the type is filed in the first memory (WM) for the type of technical system to which the technical system belongs as well as for each type of module which occurs in the technical system, that

   in the first memory (WM) for the type of technical systems to which the technical system belongs as well as for each type of modules which occurs at least once in the technical system there is filed

   o which disturbances at a technical system or each module of the type have a direct effect outwards,

   o which disturbances at a technical system or each module of the type are caused directly from outside,

   o which disturbances at a technical system or each module of the type act internally,

   o which disturbances at a technical system or each module of the type have the consequence of which other disturbances at the technical system or the module

   o and, in the case of a module of the type, whether it is elemental or composite,

   in the first memory (WM) for the type of technical systems to which the technical system belongs, as well as for each type of technical composite modules which occurs at least once in the technical system, there is filed

   o which roles must be cast in the technical system or in the module in order that the technical system or the module can fulfil its target function and

   o it is filed for each of these roles in the technical system or in the module

   - which disturbances at the role have a direct effect outwards,

   - which disturbances at the role are caused directly from outside,

   - for each disturbance St at the role having a direct effect outwards, which other disturbances at a technical system or a module of the type and which disturbances at other roles in this technical system or in this module are the consequence of the disturbance St and

   - for each disturbance St at the role caused directly from outside, which other disturbances at a technical system or a module of the type and which disturbances at other roles in this technical system or in this module are the consequence of the disturbance St.
3. Equipment according to claim 1 or 2,
   characterised thereby, that
   in the second memory (Konf),

   o there is filed which type the technical system is,

   o there is filed which type the module is, which is cast into the role, for the technical system and for each composite module in the technical system in the technical system.

   o and it is made recognisable by identifications or in other ways which modules are cast in several roles at the same time.
4. Equipment according to one of the claims 1 to 3,
   characterised thereby, that

   a remedy for each disturbance, which occurs at a technical system of the type, to which the technical system belongs, and which is caused directly from outside, is filed in the first memory (WM),

   in the first memory (WM), there is filed for the type of technical system to which the technical system belongs as well as for each type of modules which occurs at least once in the technical system

   o a remedy at the technical system or module for each disturbance at a technical system or a module of the type, which (disturbance) is either the consequence of another disturbance at the technical system or at the module or is still caused directly from outside at a role in the technical system or module,

   o which examinations can be performed at a technical system or a module of the type in the course of a diagnosis,

   o which result the examination has when the technical system is undisturbed, i.e. the target result for each examination at a technical system or a module of the type,

   o which results the examination can have apart from its target result, i.e. all the off-target results for each examination at a technical system or a module of the type,

   o the disturbances at the technical system or module, by which (disturbances) each off-target result of an examination at a technical system or a module of the type can be explained, and

   o which disturbances at the technical system or module are incompatible with each result of an examination at a technical system or a module of the type.
5. Equipment according to one of the claims 1 to 4,
   characterised thereby, that

   an unambiguous identification of the type is filed in the first memory (WM) for the type, to which the technical system belongs, as well as for each type of module which occurs in the technical system and

   in the second memory (Konf), there is filed

   o a respective unambiguous identification for the technical system and for each module in the technical system and

   o which identification the module has, which is cast in the role, for each role in the technical system and for each role in each composite module in the technical system.
6. Equipment according to one of the claims 1 to 4,
   characterised thereby, that

   the first computer (Konf-R) produces an unambiguous identification for the technical system and for each module in the technical system and

   the first computer (Konf-R), when a module is put together, ascertains the identification of the casting of the role in the module for each role in this module.
7. Equipment according to one of the claims 1 to 6,

   consisting of a first computer (Wb compiler) with reading access to a first memory (WM) and a second memory (Konf) as well as at least temporarily with writing access to a third memory (Wb-zw) and

   of a second computer (Wb-extractor) at least temporarily with reading access to the third memory (Wb-zw) and at least temporarily with writing access to a fourth memory (Wb),

   characterised thereby, that

   in a first phase the first computer (Wb compiler) compiles the knowledge base for a technical system according to one of claims 1 to 5 and files it in the third memory (Wb-zw), and

   the second computer (Wb-extractor) in a second phase

   o ascertains which single causes can occur at the technical system and its modules,

   o ascertains for each of these causes by which remedy the cause is removed,

   o ascertains which examinations can be performed at the technical system and its modules in the course of a diagnosis,

   o ascertains for each of these examinations which target result and which off-target results the examination has,

   o ascertains for each result the causes with which it is incompatible,

   o ascertains additionally for each off-target the causes by which it is explained,

   o and files all these information data in Wb.
8. Equipment according to one of the claims 1 to 7,
   characterised thereby, that
   the information data are ordered taxonomically in the first memory (WM) in that it can be stored in the description of a module type in the first memory (WM) the description of which other module type the information data about the module type can be derived when these data are absent in the description of the module type.
9. Equipment according to one of the claims 1 to 8,

   consisting of a first computer (Konf-R) with reading access to first memory (WM) and a second memory (Konf-Info) and at least temporarily with writing access to a third memory (Konf)

   and of a second computer (WB compiler) with reading access to a first memory (WM), reading access to a third memory (Konf) and at least temporarily with writing access to a fourth memory (Wb),

   characterised thereby, that

   in the second memory (Konf-Info), there is filed for the type of technical systems,to which the technical system belongs as well as for each type of modules, which occurs at least once in the technical system and for each role in a technical system or module of this type,

   o the module type of which the casting of the role is, and

   o there can be filed which other roles in the technical system or in the module are additionally occupied by a casting of this role, wherein this determination may be linked to a condition and

   o there can be filed which roles in other composite modules are additionally occupied by a casting of this role, wherein this determination may be linked to a condition and

   and the first computer (Wb compiler) in a first phase and after reading access to the second computer (Konf-Info) ascertains the configuration of the technical system and files it in the third memory (Konf)

   and thereafter produces the knowledge base for the technical system according to one of the claims 1 to 8 and files it in the fourth memory (Wb).
10. Equipment for computer-aided diagnosis of a technical system consisting of modules,

    the equipment consisting of a part equipment (e.g. Wb-compiler), with reading access to a first memory (WM) and a second memory (Konf) as well as at least temporarily with writing access to a third memory (Wb),

    and of a third computer (Diag) at least temporarily with reading access to the third memory (Wb),

    characterised thereby, that

    the part equipment automatically compiles a knowledge base for the technical system according to one of the claims 1 to 9 and files it in the third memory (Wb) in a first phase

    and the third computer (Diag) diagnoses the technical system in a second phase.
11. Equipment according to claim 10,

    the equipment consisting of a part equipment (e.g. Wb-compiler), with reading access to a first memory (WM) and a second memory (Konf) as well as at least temporarily with writing access to a third memory (Wb),

    and of a third computer (Diag) at least temporarily with reading access to the third memory (Wb) and reading and writing access to a working memory (AS),

    characterised thereby, that
    the part equipment (Wb-compiler) automatically compiles the knowledge base for a technical system according to one of the claims 1 to 9 and files it in the third memory (Wb) in a first phase and the third computer (Diag) diagnoses the technical system in a second phase in that the third computer (Diag)

    o for each reading access to the third memory (Wb) ascertains which causes can occur at the technical system and enters these causes into the working memory (AS),

    o thereafter carries out a sequence in a loop for as long as the working memory (AS) still contains several causes, which sequence consists in that the third computer (Diag)

    - automatically carries out an examination or lets it be carried out by a human being,

    - thereafter ascertains the result of the examination

    - deletes all causes from the working memory (AS), which are incompatible with the result

    - thereafter when the result is an off-target result of the examination,

    - deletes all causes from the working memory (AS), which do not explain the result,

    - and augments the working memory (AS) by all multiple causes which consist of a cause explained by the result and of a cause which the third computer (Diag) has just deleted from the working memory (AS),

    - when a single cause is still contained in the working memory (AS),

    - marks this cause as actually present and

    - automatically performs the associated remedy

    - or, when no cause is contained in the working memory (AS), marks the technical system as free of disturbance.
12. Equipment according to one of the claims 10 and 11, characterised thereby, that the second memory (Konf) is contained in the technical system to be diagnosed and the first computer (Wb-compiler) gets reading access to the second memory (Konf) by on-line access or by way of data carrier only at the time, at which the technical system is to be diagnosed and automatically produces a knowledge base.
13. Equipment according to one of the claims 9 to 12, characterised thereby, that the first memory (WM) is contained wholly or partly in the technical system to be diagnosed and the first computer (Wb-compiler) gets reading access to the second memory (Konf) by on-line access or by way of data carrier only at the time, at which the technical system is to be diagnosed and automatically produces a knowledge base.
14. Equipment according to one of the claims 9 to 13, characterised thereby, that the first computer (Konf-R) and/or the second computer (Wb-compiler) is or are identical with the third computer (Diag).
15. Equipment according to one of the claims 9 to 14, characterised thereby, that the first computer (Konf-R) and/or the second computer (Wb-compiler) and/or the third computer (Diag) is or are contained in the technical system to be diagnosed.
16. Equipment according to one of the claims 1 to 15, characterised thereby, that a characterisation of frequency of occurrence is filed for each cause which can occur at the technical system and a characterisation of the costs of performance is filed for each examination which can be performed at the technical system.
17. Equipment according to one of the claims 1 to 16, with a third computer (Diag) at least temporarily with reading access to the third memory (Wb), characterised thereby, that the third computer (Diag) in dependence on the characterisations of the frequency of occurrence of all causes and the costs of performance of all examinations selects that examination which displays the most favourable ratio of benefit to be expected for differentiation between the still possible causes and costs.
18. Equipment for the compilation of a complete disturbance graph for a technical system consisting of modules, the equipment consisting of a computer (Wb compiler) with reading access to a first memory (WM) and a second memory (Konf) as well as at least temporarily with writing access to a third memory (Wb), characterised thereby, that all information data are filed according to one of the claims 1 to 17 in the first memory (WM) and the computer (Wb-compiler) automatically compiles a complete disturbance graph for the technical system according to one of the claims 1 to 17 and files it in the third memory (Wb).

Images